Create RAID Arrays with mdadm on Ubuntu

mdadm is the standard Linux utility for creating and managing software RAID arrays. It lets you combine multiple raw block devices into a single logical device with configurable performance, redundancy, or both, without requiring dedicated hardware.

This guide walks through creating RAID 0, 1, 5, 6, and 10 arrays on Ubuntu using mdadm. For each level you will identify component disks, create the array, format and mount it, and persist the configuration across reboots. A dedicated section covers array monitoring, disk failure handling, failed disk replacement, and a comparison of mdadm software RAID against ZFS and LVM.

• RAID 0 maximises capacity and throughput but provides no redundancy.
• RAID 1 mirrors data across two disks; one disk can fail without data loss.
• RAID 5 distributes parity across three or more disks; one disk can fail.
• RAID 6 adds a second parity block; two disks can fail simultaneously.
• RAID 10 combines striping and mirroring for performance and redundancy.
• Persisting an array requires three steps: update /etc/mdadm/mdadm.conf, run update-initramfs -u, and add the device to /etc/fstab.

To follow the steps in this guide, you will need:

• A non-root user with sudo privileges on an Ubuntu server. To learn how to set up an account with these privileges, follow our Ubuntu initial server setup guide.
• A basic understanding of RAID terminology and concepts. To learn more about RAID and what RAID level is right for you, read Introduction to RAID.
• Multiple raw storage devices available on your server. The examples in this tutorial demonstrate how to configure various types of arrays on the server. As such, you will need some drives to configure.
• Depending on the array type, you will need two to four storage devices. These drives do not need to be formatted prior to following this guide.

You can skip this section for now if you have not yet set up any arrays. This guide will introduce a number of different RAID levels. If you wish to follow along and complete each RAID level for your devices, you will likely want to reuse your storage devices after each section. This specific section Resetting Existing RAID Devices can be referenced to reset your component storage devices prior to testing a new RAID level.

Begin by finding the active arrays in the /proc/mdstat file:

cat /proc/mdstat

Output
Personalities : [raid0] [linear] [multipath] [raid1] [raid6] [raid5] [raid4] [raid10]
md0 : active raid0 sdc[1] sdd[0]
      209584128 blocks super 1.2 512k chunks

            unused devices: <none>

Then unmount the array from the filesystem:

sudo umount /dev/md0

Now stop and remove the array:

sudo mdadm --stop /dev/md0

Find the devices that were used to build the array with the following command:

lsblk -o NAME,SIZE,FSTYPE,TYPE,MOUNTPOINT

Output
NAME     SIZE FSTYPE            TYPE MOUNTPOINT
sda      100G linux_raid_member disk 
sdb      100G linux_raid_member disk 
sdc      100G                   disk
sdd      100G                   disk
vda       25G                   disk
├─vda1  24.9G ext4              part /
├─vda14    4M                   part
└─vda15  106M vfat              part /boot/efi
vdb      466K iso9660           disk

The linux_raid_member value in the FSTYPE column identifies disks that are currently part of a RAID array. After identifying them, zero their superblock, which holds the RAID metadata. Zeroing the superblock removes that metadata and returns the disk to a plain, unformatted state:

sudo mdadm --zero-superblock /dev/sda
sudo mdadm --zero-superblock /dev/sdb

It’s recommended to also remove any persistent references to the array. Edit the /etc/fstab file and comment out or remove the reference to your array. You can comment it out by inserting a hashtag symbol # at the beginning of the line, using nano or your preferred text editor:

sudo nano /etc/fstab

. . .
# /dev/md0 /mnt/md0 ext4 defaults,nofail,discard 0 0

Also, comment out or remove the array definition from the /etc/mdadm/mdadm.conf file:

sudo nano /etc/mdadm/mdadm.conf

. . .
# ARRAY /dev/md0 metadata=1.2 name=mdadmwrite:0 UUID=7261fb9c:976d0d97:30bc63ce:85e76e91

Finally, update the initramfs again so that the early boot process does not try to bring an unavailable array online:

sudo update-initramfs -u

From here, you should be ready to reuse the storage devices individually, or as components of a different array.

The RAID 0 array works by breaking up data into chunks and striping it across the available disks. This means that each disk contains a portion of the data and that multiple disks will be referenced when retrieving information.

• Requirements: Minimum of 2 storage devices.
• Primary benefit: Full aggregate capacity and maximum sequential throughput. Two 100G disks give you a single 200G device with roughly double the read/write bandwidth of a single disk.
• When to use it: Scratch disks, video editing working storage, caching layers, or any workload where speed matters and data is either disposable or backed up elsewhere.
• Critical limitation: A single disk failure destroys the entire array with no recovery path. Do not use RAID 0 for any data you cannot afford to lose. Always maintain an independent backup.

Identifying the Component Devices (RAID 0)

To start, find the identifiers for the raw disks that you will be using:

lsblk -o NAME,SIZE,FSTYPE,TYPE,MOUNTPOINT

Output
NAME     SIZE FSTYPE   TYPE MOUNTPOINT
sda      100G          disk
sdb      100G          disk
vda       25G          disk
├─vda1  24.9G ext4     part /
├─vda14    4M          part
└─vda15  106M vfat     part /boot/efi
vdb      466K iso9660  disk

In this example, you have two disks without a filesystem, each 100G in size. These devices have been given the /dev/sda and /dev/sdb identifiers for this session and will be the raw components used to build the array. The vda device is the system disk. The vdb entry with iso9660 filesystem is a read-only cloud-init or metadata disk attached by the hypervisor — you can ignore it. You are only working with the raw, unformatted disks, which are the ones showing no FSTYPE value.

Creating the RAID 0 Array

To create a RAID 0 array with these components, pass them into the mdadm --create command. You will have to specify the device name you wish to create, the RAID level, and the number of devices. In this command example, you will be naming the device /dev/md0, and include the two disks that will build the array:

sudo mdadm --create --verbose /dev/md0 --level=0 --raid-devices=2 /dev/sda /dev/sdb

Output
mdadm: layout defaults to -p original
mdadm: chunk size defaults to 512K
mdadm: size set to 104792064K
mdadm: Defaulting to version 1.2 metadata
mdadm: array /dev/md0 started.

Confirm that the RAID was successfully created by checking the /proc/mdstat file:

cat /proc/mdstat

Output
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md0 : active raid0 sdb[1] sda[0]
      209584128 blocks super 1.2 512k chunks

            unused devices: <none>

This output reveals that the /dev/md0 device was created in the RAID 0 configuration using the /dev/sda and /dev/sdb devices.

Creating and Mounting the Filesystem (RAID 0)

Next, create an ext4 filesystem on the array. The -F flag forces mkfs.ext4 to proceed even though /dev/md0 is not a partition — it is a whole block device, and mkfs requires this flag to confirm you intend to format the entire device:

sudo mkfs.ext4 -F /dev/md0

Then, create a mount point to attach the new filesystem:

sudo mkdir -p /mnt/md0

You can mount the filesystem with the following command:

sudo mount /dev/md0 /mnt/md0

After mounting, check whether the new space is available:

df -h -x devtmpfs -x tmpfs

Output
Filesystem      Size  Used Avail Use% Mounted on
/dev/vda1        25G  1.4G   23G   6% /
/dev/vda15      105M  3.4M  102M   4% /boot/efi
/dev/md0        196G   61M  186G   1% /mnt/md0

The new filesystem is mounted and accessible. The available space shows 196G rather than the full 200G (2x 100G) because ext4 reserves approximately 5% of the filesystem for root and internal metadata by default.

Saving the RAID 0 Array Layout

Persisting the array across reboots requires three separate steps, each solving a different failure mode:

1. mdadm --detail --scan writes the array definition including its UUID to /etc/mdadm/mdadm.conf, which is what mdadm reads at boot to know which arrays to assemble and under which device names.
2. update-initramfs -u rebuilds the early boot image to include the updated mdadm.conf. Without this step, the initramfs used during boot still reflects the old configuration and the array may not assemble at all, or may assemble under an unexpected name like /dev/md127.
3. The /etc/fstab entry with nofail tells the OS where to mount the array after it assembles. The nofail flag is critical: if the array fails to assemble at boot for any reason, nofail allows the system to continue booting instead of dropping to emergency mode.

Skipping any one of these three steps is the most common cause of an array that works on first boot but fails to come back after a reboot.

sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf

sudo update-initramfs -u

Add the new filesystem mount options to the /etc/fstab file for automatic mounting at boot:

echo '/dev/md0 /mnt/md0 ext4 defaults,nofail,discard 0 0' | sudo tee -a /etc/fstab

Your RAID 0 array will now automatically assemble and mount each boot.

You’re now finished with your RAID set up. If you want to try a different RAID, follow the resetting instructions at the beginning of this tutorial to proceed with creating a new RAID array type.

The RAID 1 array type is implemented by mirroring data across all available disks. Each disk in a RAID 1 array gets a full copy of the data, providing redundancy in the event of a device failure.

• Requirements: Minimum of 2 storage devices.
• Primary benefit: Full mirror redundancy. Either disk can fail completely and the array keeps running without data loss. Read performance can improve because mdadm can service read requests from both disks simultaneously.
• When to use it: OS disks, boot volumes, and any two-disk setup where uptime matters more than capacity. RAID 1 is the standard choice for the system disk on a production server.
• Capacity tradeoff: Usable space equals the size of the smallest disk. Two 100G disks give you 100G of usable storage.

Identifying the Component Devices (RAID 1)

To start, find the identifiers for the raw disks that you will be using:

lsblk -o NAME,SIZE,FSTYPE,TYPE,MOUNTPOINT

Output
NAME     SIZE FSTYPE   TYPE MOUNTPOINT
sda      100G          disk 
sdb      100G          disk 
vda       25G          disk
├─vda1  24.9G ext4     part /
├─vda14    4M          part
└─vda15  106M vfat     part /boot/efi
vdb      466K iso9660  disk

In this example, you have two disks without a filesystem, each 100G in size. These devices have been given the /dev/sda and /dev/sdb identifiers for this session and will be the raw components you use to build the array.

Creating the RAID 1 Array

To create a RAID 1 array with these components, pass them into the mdadm --create command. You will have to specify the device name you wish to create, the RAID level, and the number of devices. In this command example, you will be naming the device /dev/md0, and include the disks that will build the array:

sudo mdadm --create --verbose /dev/md0 --level=1 --raid-devices=2 /dev/sda /dev/sdb

If the component devices you are using are not partitions with the boot flag enabled, you will likely receive the following warning. It is safe to respond with y and continue:

Output
mdadm: Note: this array has metadata at the start and
    may not be suitable as a boot device.  If you plan to
    store '/boot' on this device please ensure that
    your boot-loader understands md/v1.x metadata, or use
    --metadata=0.90
mdadm: size set to 104792064K
Continue creating array? y

The mdadm tool will start to mirror the drives. This can take some time to complete, but the array can be used during this time. You can monitor the progress of the mirroring by checking the /proc/mdstat file:

cat /proc/mdstat

Output
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md0 : active raid1 sdb[1] sda[0]
      104792064 blocks super 1.2 [2/2] [UU]
      [====>................]  resync = 20.2% (21233216/104792064) finish=6.9min speed=199507K/sec

unused devices: <none>

In the first highlighted line, the /dev/md0 device was created in the RAID 1 configuration using the /dev/sda and /dev/sdb devices. The second highlighted line reveals the progress on the mirroring. You can continue to the next step while this process completes.

Creating and Mounting the Filesystem (RAID 1)

Next, create an ext4 filesystem on the array. The -F flag forces mkfs.ext4 to proceed even though /dev/md0 is not a partition — it is a whole block device, and mkfs requires this flag to confirm you intend to format the entire device:

sudo mkfs.ext4 -F /dev/md0

Then, create a mount point to attach the new filesystem:

sudo mkdir -p /mnt/md0

You can mount the filesystem by running the following:

sudo mount /dev/md0 /mnt/md0

Check whether the new space is available:

df -h -x devtmpfs -x tmpfs

Output
Filesystem      Size  Used Avail Use% Mounted on
/dev/vda1        25G  1.4G   23G   6% /
/dev/vda15      105M  3.4M  102M   4% /boot/efi
/dev/md0         99G   60M   94G   1% /mnt/md0

The new filesystem is mounted and accessible. The available space shows 94G rather than the full 100G because ext4 reserves approximately 5% of the filesystem for root and internal metadata by default.

Saving the RAID 1 Array Layout

As with RAID 0, persisting this array across reboots requires the same three steps, each solving a different failure mode:

1. mdadm --detail --scan writes the array definition including its UUID to /etc/mdadm/mdadm.conf, which is what mdadm reads at boot to know which arrays to assemble and under which device names.
2. update-initramfs -u rebuilds the early boot image to include the updated mdadm.conf. Without this step, the initramfs used during boot still reflects the old configuration and the array may not assemble at all, or may assemble under an unexpected name like /dev/md127.
3. The /etc/fstab entry with nofail tells the OS where to mount the array after it assembles. The nofail flag is critical: if the array fails to assemble at boot for any reason, nofail allows the system to continue booting instead of dropping to emergency mode.

Skipping any one of these three steps is the most common cause of an array that works on first boot but fails to come back after a reboot.

sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf

sudo update-initramfs -u

Add the new filesystem mount options to the /etc/fstab file for automatic mounting at boot:

echo '/dev/md0 /mnt/md0 ext4 defaults,nofail,discard 0 0' | sudo tee -a /etc/fstab

Your RAID 1 array will now automatically assemble and mount each boot.

You’re now finished with your RAID set up. If you want to try a different RAID, follow the resetting instructions at the beginning of this tutorial to proceed with creating a new RAID array type.

The RAID 5 array type is implemented by striping data across the available devices. One component of each stripe is a calculated parity block. If a device fails, the parity block and the remaining blocks can be used to calculate the missing data. The device that receives the parity block is rotated so that each device has a balanced amount of parity information.

• Requirements: Minimum of 3 storage devices.
• Primary benefit: Distributed parity across all member disks. One disk can fail and the array remains fully readable using parity reconstruction.
• When to use it: General-purpose file storage where you need more capacity than RAID 1 but still need one disk’s worth of fault tolerance. With N disks you get N-1 disks of usable space.
• Production caveat: RAID 5 rebuild times on large disks can run several hours. During a rebuild, a second disk failure destroys the array. On disks larger than 2TB, seriously consider RAID 6 instead. RAID 5 also has significantly degraded write performance while in degraded state.

Identifying the Component Devices (RAID 5)

To start, find the identifiers for the raw disks that you will be using:

lsblk -o NAME,SIZE,FSTYPE,TYPE,MOUNTPOINT

Output
NAME     SIZE FSTYPE   TYPE MOUNTPOINT
sda      100G          disk 
sdb      100G          disk 
sdc      100G          disk 
vda       25G          disk
├─vda1  24.9G ext4     part /
├─vda14    4M          part
└─vda15  106M vfat     part /boot/efi
vdb      466K iso9660  disk

You have three disks without a filesystem, each 100G in size. These devices have been given the /dev/sda, /dev/sdb, and /dev/sdc identifiers for this session and will be the raw components you use to build the array.

Creating the RAID 5 Array

To create a RAID 5 array with these components, pass them into the mdadm --create command. You will have to specify the device name you wish to create, the RAID level, and the number of devices. In this command example, you will be naming the device /dev/md0, and include the disks that will build the array:

sudo mdadm --create --verbose /dev/md0 --level=5 --raid-devices=3 /dev/sda /dev/sdb /dev/sdc

Output
mdadm: layout defaults to left-symmetric
mdadm: chunk size defaults to 512K
mdadm: size set to 104791040K
mdadm: Defaulting to version 1.2 metadata
mdadm: array /dev/md0 started.

The mdadm tool will start to configure the array. It uses the recovery process to build the array because writing parity incrementally as data is placed is faster than zeroing all blocks first — this means the array is technically in a recovery state during the initial build. This can take some time to complete, but the array can be used during this time. You can monitor the progress of the mirroring by checking the /proc/mdstat file:

cat /proc/mdstat

Output
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md0 : active raid5 sdc[3] sdb[1] sda[0]
      209582080 blocks super 1.2 level 5, 512k chunk, algorithm 2 [3/2] [UU_]
      [>....................]  recovery =  0.9% (957244/104791040) finish=18.0min speed=95724K/sec

unused devices: <none>

In the first highlighted line, the /dev/md0 device was created in the RAID 5 configuration using the /dev/sda, /dev/sdb and /dev/sdc devices. The second highlighted line shows the progress of the build.

You can continue the guide while this process completes.

Creating and Mounting the Filesystem (RAID 5)

Next, create an ext4 filesystem on the array. The -F flag forces mkfs.ext4 to proceed even though /dev/md0 is not a partition — it is a whole block device, and mkfs requires this flag to confirm you intend to format the entire device:

sudo mkfs.ext4 -F /dev/md0

Create a mount point to attach the new filesystem:

sudo mkdir -p /mnt/md0

You can mount the filesystem with the following:

sudo mount /dev/md0 /mnt/md0

Check whether the new space is available:

df -h -x devtmpfs -x tmpfs

Output
Filesystem      Size  Used Avail Use% Mounted on
/dev/vda1        25G  1.4G   23G   6% /
/dev/vda15      105M  3.4M  102M   4% /boot/efi
/dev/md0        197G   60M  187G   1% /mnt/md0

The new filesystem is mounted and accessible.

Saving the RAID 5 Array Layout

As with RAID 0, persisting this array across reboots requires the same three steps, each solving a different failure mode:

1. mdadm --detail --scan writes the array definition including its UUID to /etc/mdadm/mdadm.conf, which is what mdadm reads at boot to know which arrays to assemble and under which device names.
2. update-initramfs -u rebuilds the early boot image to include the updated mdadm.conf. Without this step, the initramfs used during boot still reflects the old configuration and the array may not assemble at all, or may assemble under an unexpected name like /dev/md127.
3. The /etc/fstab entry with nofail tells the OS where to mount the array after it assembles. The nofail flag is critical: if the array fails to assemble at boot for any reason, nofail allows the system to continue booting instead of dropping to emergency mode.

Skipping any one of these three steps is the most common cause of an array that works on first boot but fails to come back after a reboot.

You can monitor the progress of the mirroring by checking the /proc/mdstat file:

cat /proc/mdstat

Output
Personalities : [raid1] [linear] [multipath] [raid0] [raid6] [raid5] [raid4] [raid10]
md0 : active raid5 sdc[3] sdb[1] sda[0]
      209584128 blocks super 1.2 level 5, 512k chunk, algorithm 2 [3/3] [UUU]

unused devices: <none>

This output reveals that the rebuild is complete. Now, you can automatically scan the active array and append the file:

sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf

sudo update-initramfs -u

Add the new filesystem mount options to the /etc/fstab file for automatic mounting at boot:

echo '/dev/md0 /mnt/md0 ext4 defaults,nofail,discard 0 0' | sudo tee -a /etc/fstab

Your RAID 5 array will now automatically assemble and mount each boot.

You’re now finished with your RAID set up. If you want to try a different RAID, follow the resetting instructions at the beginning of this tutorial to proceed with creating a new RAID array type.

The RAID 6 array type is implemented by striping data across the available devices. Two components of each stripe are calculated parity blocks. If one or two devices fail, the parity blocks and the remaining blocks can be used to calculate the missing data. The devices that receive the parity blocks are rotated so that each device has a balanced amount of parity information. This is similar to a RAID 5 array, but allows for the failure of two drives.

• Requirements: Minimum of 4 storage devices.
• Primary benefit: Double parity. Two disks can fail simultaneously and the array continues to operate. This makes RAID 6 the safer choice for large-disk arrays where a simultaneous second failure during a long rebuild is a realistic risk.
• When to use it: High-capacity storage arrays with 4 or more large disks. The two-disk failure tolerance provides a meaningful safety margin that RAID 5 cannot offer.
• Capacity tradeoff: Two disks worth of capacity is used for parity. Four 100G disks give you 200G of usable storage.

Identifying the Component Devices (RAID 6)

To start, find the identifiers for the raw disks that you will be using:

lsblk -o NAME,SIZE,FSTYPE,TYPE,MOUNTPOINT

Output
NAME     SIZE FSTYPE   TYPE MOUNTPOINT
sda      100G          disk 
sdb      100G          disk 
sdc      100G          disk 
sdd      100G          disk 
vda       25G          disk
├─vda1  24.9G ext4     part /
├─vda14    4M          part
└─vda15  106M vfat     part /boot/efi
vdb      466K iso9660  disk

In this example, you have four disks without a filesystem, each 100G in size. These devices have been given the /dev/sda, /dev/sdb, /dev/sdc, and /dev/sdd identifiers for this session and will be the raw components used to build the array.

Creating the RAID 6 Array

To create a RAID 6 array with these components, pass them into the mdadm --create command. You have to specify the device name you wish to create, the RAID level, and the number of devices. In this following command example, you will be naming the device /dev/md0 and include the disks that will build the array:

sudo mdadm --create --verbose /dev/md0 --level=6 --raid-devices=4 /dev/sda /dev/sdb /dev/sdc /dev/sdd

Output
mdadm: layout defaults to left-symmetric
mdadm: chunk size defaults to 512K
mdadm: size set to 104792064K
mdadm: Defaulting to version 1.2 metadata
mdadm: array /dev/md0 started.

The mdadm tool will start to configure the array. It uses the recovery process to build the array because writing parity incrementally as data is placed is faster than zeroing all blocks first — this means the array is technically in a recovery state during the initial build. This can take some time to complete, but the array can be used during this time. You can monitor the progress of the mirroring by checking the /proc/mdstat file:

cat /proc/mdstat

Output
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md0 : active raid6 sdd[3] sdc[2] sdb[1] sda[0]
      209584128 blocks super 1.2 level 6, 512k chunk, algorithm 2 [4/4] [UUUU]
      [>....................]  resync =  0.6% (668572/104792064) finish=10.3min speed=167143K/sec

unused devices: <none>

In the first highlighted line, the /dev/md0 device has been created in the RAID 6 configuration using the /dev/sda, /dev/sdb, /dev/sdc and /dev/sdd devices. The second highlighted line shows the progress of the build. You can continue the guide while this process completes.

Creating and Mounting the Filesystem (RAID 6)

Next, create an ext4 filesystem on the array. The -F flag forces mkfs.ext4 to proceed even though /dev/md0 is not a partition — it is a whole block device, and mkfs requires this flag to confirm you intend to format the entire device:

sudo mkfs.ext4 -F /dev/md0

Create a mount point to attach the new filesystem:

sudo mkdir -p /mnt/md0

You can mount the filesystem with the following:

sudo mount /dev/md0 /mnt/md0

Check whether the new space is available:

df -h -x devtmpfs -x tmpfs

Output
Filesystem      Size  Used Avail Use% Mounted on
/dev/vda1        25G  1.4G   23G   6% /
/dev/vda15      105M  3.4M  102M   4% /boot/efi
/dev/md0        197G   60M  187G   1% /mnt/md0

The new filesystem is mounted and accessible.

Saving the RAID 6 Array Layout

As with RAID 0, persisting this array across reboots requires the same three steps, each solving a different failure mode:

1. mdadm --detail --scan writes the array definition including its UUID to /etc/mdadm/mdadm.conf, which is what mdadm reads at boot to know which arrays to assemble and under which device names.
2. update-initramfs -u rebuilds the early boot image to include the updated mdadm.conf. Without this step, the initramfs used during boot still reflects the old configuration and the array may not assemble at all, or may assemble under an unexpected name like /dev/md127.
3. The /etc/fstab entry with nofail tells the OS where to mount the array after it assembles. The nofail flag is critical: if the array fails to assemble at boot for any reason, nofail allows the system to continue booting instead of dropping to emergency mode.

Skipping any one of these three steps is the most common cause of an array that works on first boot but fails to come back after a reboot.

sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf

sudo update-initramfs -u

Add the new filesystem mount options to the /etc/fstab file for automatic mounting at boot:

echo '/dev/md0 /mnt/md0 ext4 defaults,nofail,discard 0 0' | sudo tee -a /etc/fstab

Your RAID 6 array will now automatically assemble and mount each boot.

You’re now finished with your RAID set up. If you want to try a different RAID, follow the resetting instructions at the beginning of this tutorial to proceed with creating a new RAID array type.

The RAID 10 array type is traditionally implemented by creating a striped RAID 0 array composed of sets of RAID 1 arrays. This nested array type gives both redundancy and high performance, at the expense of large amounts of disk space. The mdadm utility has its own RAID 10 type that provides the same type of benefits with increased flexibility. It is not created by nesting arrays, but has many of the same characteristics and guarantees. You will be using the mdadm RAID 10 here.

• Requirements: Minimum of 4 storage devices for a standard setup. mdadm’s native RAID 10 technically accepts 3 disks, but 4 is the practical minimum for balanced mirroring.
• Primary benefit: Combines RAID 0 striping with RAID 1 mirroring. You get both the throughput of striping and the redundancy of mirroring. One disk per mirrored pair can fail without data loss.
• When to use it: Database servers, high-throughput application storage, or any workload that needs both performance and redundancy. RAID 10 has the fastest rebuild time of any redundant RAID level because recovery only involves copying data from the surviving mirror, not computing parity.
• Capacity tradeoff: Half of total disk capacity is used for mirroring. Four 100G disks give you 200G of usable storage.

By default, two copies of each data block will be stored in what is called the near layout. The possible layouts that dictate how each data block is stored are as follows:

• near: The default arrangement. Copies of each chunk are written consecutively when striping, meaning that the copies of the data blocks will be written around the same part of multiple disks.
• far: The first and subsequent copies are written to different parts of the storage devices in the array. For instance, the first chunk might be written near the beginning of a disk, while the second chunk would be written halfway down on a different disk. This can give some read performance gains for traditional spinning disks at the expense of write performance.
• offset: Each stripe is copied, and offset by one drive. This means that the copies are offset from one another, but still close together on the disk. This helps minimize excessive seeking during some workloads.

You can find out more about these layouts by checking out the RAID10 section of this man page:

man 4 md

Identifying the Component Devices (RAID 10)

To start, find the identifiers for the raw disks that you will be using:

lsblk -o NAME,SIZE,FSTYPE,TYPE,MOUNTPOINT

Output
NAME     SIZE FSTYPE   TYPE MOUNTPOINT
sda      100G          disk 
sdb      100G          disk 
sdc      100G          disk 
sdd      100G          disk 
vda       25G          disk
├─vda1  24.9G ext4     part /
├─vda14    4M          part
└─vda15  106M vfat     part /boot/efi
vdb      466K iso9660  disk

In this example, you have four disks without a filesystem, each 100G in size. These devices have been given the /dev/sda, /dev/sdb, /dev/sdc, and /dev/sdd identifiers for this session and will be the raw components used to build the array.

Creating the RAID 10 Array

To create a RAID 10 array with these components, pass them into the mdadm --create command. You have to specify the device name you wish to create, the RAID level, and the number of devices. In this following command example, you will be naming the device /dev/md0 and include the disks that will build the array:

You can set up two copies using the near layout by not specifying a layout and copy number:

sudo mdadm --create --verbose /dev/md0 --level=10 --raid-devices=4 /dev/sda /dev/sdb /dev/sdc /dev/sdd

Output
mdadm: layout defaults to n2
mdadm: chunk size defaults to 512K
mdadm: size set to 104792064K
mdadm: Defaulting to version 1.2 metadata
mdadm: array /dev/md0 started.

If you want to use a different layout or change the number of copies, you will have to use the --layout= option, which takes a layout and copy identifier. The layouts are n for near, f for far, and o for offset. The number of copies to store is appended afterward.

For instance, to create an array that has three copies in the offset layout, the command would include the following:

sudo mdadm --create --verbose /dev/md0 --level=10 --layout=o3 --raid-devices=4 /dev/sda /dev/sdb /dev/sdc /dev/sdd

The mdadm tool will start to configure the array. It uses the recovery process to build the array because writing parity incrementally as data is placed is faster than zeroing all blocks first — this means the array is technically in a recovery state during the initial build. This can take some time to complete, but the array can be used during this time. You can monitor the progress of the mirroring by checking the /proc/mdstat file:

cat /proc/mdstat

Output
Personalities : [raid6] [raid5] [raid4] [linear] [multipath] [raid0] [raid1] [raid10]
md0 : active raid10 sdd[3] sdc[2] sdb[1] sda[0]
      209584128 blocks super 1.2 512K chunks 2 near-copies [4/4] [UUUU]
      [===>.................]  resync = 18.1% (37959424/209584128) finish=13.8min speed=206120K/sec

unused devices: <none>

In the first highlighted line, the /dev/md0 device has been created in the RAID 10 configuration using the /dev/sda, /dev/sdb, /dev/sdc and /dev/sdd devices. The second highlighted area shows the layout that was used for this example (two copies in the near configuration). The third highlighted area shows the progress on the build. You can continue the guide while this process completes.

Creating and Mounting the Filesystem (RAID 10)

Next, create an ext4 filesystem on the array. The -F flag forces mkfs.ext4 to proceed even though /dev/md0 is not a partition — it is a whole block device, and mkfs requires this flag to confirm you intend to format the entire device:

sudo mkfs.ext4 -F /dev/md0

Create a mount point to attach the new filesystem:

sudo mkdir -p /mnt/md0

You can mount the filesystem with the following:

sudo mount /dev/md0 /mnt/md0

Check whether the new space is available:

df -h -x devtmpfs -x tmpfs

Output
Filesystem      Size  Used Avail Use% Mounted on
/dev/vda1        25G  1.4G   23G   6% /
/dev/vda15      105M  3.4M  102M   4% /boot/efi
/dev/md0        197G   60M  187G   1% /mnt/md0

The new filesystem is mounted and accessible.

Saving the RAID 10 Array Layout

As with RAID 0, persisting this array across reboots requires the same three steps, each solving a different failure mode:

1. mdadm --detail --scan writes the array definition including its UUID to /etc/mdadm/mdadm.conf, which is what mdadm reads at boot to know which arrays to assemble and under which device names.
2. update-initramfs -u rebuilds the early boot image to include the updated mdadm.conf. Without this step, the initramfs used during boot still reflects the old configuration and the array may not assemble at all, or may assemble under an unexpected name like /dev/md127.
3. The /etc/fstab entry with nofail tells the OS where to mount the array after it assembles. The nofail flag is critical: if the array fails to assemble at boot for any reason, nofail allows the system to continue booting instead of dropping to emergency mode.

Skipping any one of these three steps is the most common cause of an array that works on first boot but fails to come back after a reboot.

sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf

sudo update-initramfs -u

Add the new filesystem mount options to the /etc/fstab file for automatic mounting at boot:

echo '/dev/md0 /mnt/md0 ext4 defaults,nofail,discard 0 0' | sudo tee -a /etc/fstab

Your RAID 10 array will now automatically assemble and mount each boot.

Once an array is running, mdadm provides several tools to inspect its health, respond to disk failures, and replace degraded members.

Reading /proc/mdstat

The /proc/mdstat file is the fastest way to check array state:

cat /proc/mdstat

Output
Personalities : [raid1] [raid5] [raid10]
md0 : active raid1 sdb[1] sda[0]
      104792064 blocks super 1.2 [2/2] [UU]

unused devices: <none>

Each U in the [UU] field represents a healthy member disk. A _ indicates a missing or degraded member. For a deeper summary of a specific array, use --detail:

sudo mdadm --detail /dev/md0

Output
/dev/md0:
           Version : 1.2
     Creation Time : Mon Jan  1 00:00:00
        Raid Level : raid1
        Array Size : 104792064 (99.94 GiB 107.31 GB)
     Used Dev Size : 104792064 (99.94 GiB 107.31 GB)
      Raid Devices : 2
     Total Devices : 2
       Persistence : Superblock is persistent

       Update Time : Mon Jan  1 00:05:00
             State : clean
    Active Devices : 2
   Working Devices : 2
    Failed Devices : 0
     Spare Devices : 0

              Name : hostname:0  (local to host hostname)
              UUID : xxxxxxxx:xxxxxxxx:xxxxxxxx:xxxxxxxx
            Events : 17

    Number   Major   Minor   RaidDevice State
       0     8        0        0      active sync   /dev/sda
       1     8       16        1      active sync   /dev/sdb

Enabling Email Alerts with mdadm --monitor

Configure mdadm to send email notifications when array events occur. Add the following line to /etc/mdadm/mdadm.conf:

sudo nano /etc/mdadm/mdadm.conf

MAILADDR admin@example.com

Then enable and start the mdmonitor service:

sudo systemctl enable mdmonitor
sudo systemctl start mdmonitor

Verify the service is running:

sudo systemctl status mdmonitor

Output
● mdmonitor.service - MD array monitor
     Loaded: loaded (/lib/systemd/system/mdmonitor.service; enabled)
     Active: active (running)

mdadm --monitor sends alerts for events including Fail, DegradedArray, SparesMissing, and RebuildFinished.

Adding a Hot Spare

A hot spare is a standby disk that mdadm automatically uses to begin rebuilding the array the moment a member disk fails. To add a spare to an existing array:

sudo mdadm /dev/md0 --add /dev/sde

Output
mdadm: added /dev/sde

Confirm the spare is registered:

sudo mdadm --detail /dev/md0 | grep spare

Output
     Spare Devices : 1
       2     8       64        -      spare   /dev/sde

Handling a Failed Disk

When mdadm marks a disk as failed, the array enters a degraded state. The /proc/mdstat output will show a _ in the device status field:

Output
md0 : active raid1 sdb[1] sda[2](F)
      104792064 blocks super 1.2 [2/1] [_U]

The (F) flag on /dev/sda confirms the failure. If no hot spare is configured, the array continues to operate in degraded mode, but has no further redundancy until the failed disk is replaced.

Replacing a Failed Disk

To replace the failed disk, first mark it as failed (if mdadm has not already done so automatically):

sudo mdadm /dev/md0 --fail /dev/sda

Remove the failed disk from the array:

sudo mdadm /dev/md0 --remove /dev/sda

Physically replace the disk, then add the new disk to the array:

sudo mdadm /dev/md0 --add /dev/sda

mdadm immediately begins rebuilding the array onto the replacement disk. Monitor the rebuild progress:

watch -n 1 cat /proc/mdstat

Output
md0 : active raid1 sda[2] sdb[1]
      104792064 blocks super 1.2 [2/1] [_U]
      [=>...................]  recovery =  6.4% (6710784/104792064) finish=13.8min speed=118080K/sec

The array returns to a fully healthy state when the rebuild finishes and the status field shows [UU].

mdadm software RAID is not the only option for redundant storage on Linux. The table below summarises when to use each approach.

When to choose mdadm:

• You are on bare-metal hardware without a hardware RAID controller.
• You need a well-supported, kernel-native solution with a long track record on Ubuntu.
• You want to combine redundancy levels (RAID 10) without additional licensing concerns.

When to consider ZFS instead:

• Data integrity at scale is the primary concern (file servers, backup targets).
• You require native snapshot and send/receive capabilities.
• You can meet the memory requirements.

For volume management layered on top of an existing RAID array, see our guide on how to use LVM to manage storage devices on Ubuntu.

What is the difference between RAID 5 and RAID 10 in mdadm?

RAID 5 distributes parity across three or more disks, using one disk-equivalent of capacity for parity. It tolerates one disk failure. RAID 10 mirrors pairs of disks and stripes across the mirrors, tolerating one failure per mirrored pair. RAID 10 generally delivers better write performance and faster rebuild times; RAID 5 gives more usable capacity per disk added.

How do I check the status of a RAID array created with mdadm?

Run cat /proc/mdstat for a quick overview of all active arrays. For detailed information about a specific array including member disk state, run sudo mdadm --detail /dev/md0.

How do I make a mdadm RAID array persist after a reboot on Ubuntu?

Three steps are required. First, append the array definition to /etc/mdadm/mdadm.conf by running sudo mdadm --detail --scan | sudo tee -a /etc/mdadm/mdadm.conf. Second, update the initramfs with sudo update-initramfs -u. Third, add the mount entry to /etc/fstab using the nofail option so the system boots even if the array is unavailable.

Can I add a hot spare to an existing mdadm RAID array?

Yes. Run sudo mdadm /dev/md0 --add /dev/sdX where /dev/sdX is the spare disk. mdadm registers it as a spare and will automatically begin a rebuild onto it if a member disk fails.

What happens when a disk in a mdadm RAID array fails?

mdadm marks the disk with a (F) flag in /proc/mdstat and the array enters degraded state. If a hot spare is configured, mdadm begins rebuilding automatically. If no spare is present, the array continues operating with reduced redundancy. RAID 0 has no redundancy: a single disk failure destroys all data.

How do I replace a failed disk in a mdadm RAID 1 or RAID 5 array?

Mark the disk as failed with sudo mdadm /dev/md0 --fail /dev/sdX, remove it with sudo mdadm /dev/md0 --remove /dev/sdX, replace the physical disk, then add the new disk with sudo mdadm /dev/md0 --add /dev/sdX. The array begins rebuilding immediately.

What is the /etc/mdadm/mdadm.conf file used for?

/etc/mdadm/mdadm.conf stores persistent array definitions including the array UUID, member devices, and monitoring configuration such as the alert email address. Without a correct entry here, mdadm cannot reliably reassemble the array at boot using the expected device name.

How does mdadm software RAID compare to ZFS on Ubuntu?

mdadm is kernel-native, lightweight, and well-tested across a wide range of hardware. ZFS adds built-in data checksumming, copy-on-write snapshots, and RAID-Z (similar to RAID 5/6), but requires more RAM and has licensing constraints that prevent it from being shipped in the Linux kernel directly. For most general-purpose redundancy use cases on Ubuntu servers, mdadm is the straightforward choice. For large-scale file servers where silent data corruption is a concern, ZFS is worth the additional overhead.

In this guide, you created RAID 0, 1, 5, 6, and 10 arrays using mdadm on Ubuntu. You formatted and mounted each array, persisted the configuration across reboots by updating /etc/mdadm/mdadm.conf and running update-initramfs -u, and set up monitoring and disk replacement procedures.

As next steps:

• To manage day-to-day operations on your arrays, see the DigitalOcean guide on how to manage RAID arrays with mdadm on Ubuntu.
• To share your RAID array over the network, see how to set up an NFS mount on Ubuntu.
• For volume management on top of your array, see how to use LVM to manage storage devices on Ubuntu.